Learning user preferences in online dating
and e Harmony both use their own proprietary questionnaires that aim to dig deep into who you are, and what you may like in a partner.
The challenge in predictive modeling in dating sites is in understanding what self-reported data is “real” in the prediction models.
For example, females tend to lie about their weight, age, and build, while males tend to lie about their height, income, and age.
Another instance of providing inaccurate data is when the person believes that he/she is more appealing when listing that they love listening to classical music--while the accuracy of this data can better be determined by an analysis of the Spotify playlist or i Tunes history.
By switching to Mongo DB, they have successfully reduced the time for the compatibility matching system algorithm to run at 95% (less than 12 hours).
Big data and machine learning processes analyze a billion prospective matches a day.